Palaeogeography, Palaeoclimatology, Palaeoecology

Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213 Contents lists available at ScienceDirect Palaeogeography, Palaeoclimatology, Palaeoecology journal homepage: www.elsevier.com/locate/palaeo A 35,000 year record of changes in the eastern Indian Ocean offshore Sumatra Davide S. Murgese a,b, Patrick De Deckker a,c,, Michelle I. Spooner a,d, Martin Young a,e a Department of Earth and Marine Sciences, The Australian National University, Canberra ACT 0200, Australia b SEA Consulting S.r.l, Via Cernaia 27-10121 Torino, Italy c Research School of Earth Sciences, The Australian National University, Canberra ACT 0200, Australia d Geoscience Australia, GPO Box 378, Canberra ACT 2601, Australia e CSIRO Petroleum, 11 Julius Avenue, Riverside Corporate Park, North Ryde NSW 2113, Australia ARTICLE INFO ABSTRACT Article history: Received 14 March 2008 Received in revised form 31 May 2008 Accepted 2 June 2008 Keywords: Last glacial maximum Deep chlorophyll maximum Barrier layer Java Upwelling System Monsoon Organic carbon Foraminifera Dinoflagellates We have examined the upper 276 cm of deep-sea core BAR9403 taken at a water depth of 2034 m offshore the southern portion of Sumatra in the eastern Indian Ocean using several micropalaeontological proxies. Faunal counts and stable isotopes of oxygen and carbon of planktic and benthic foraminifers, as well as floral counts of dinoflagellates were obtained to reconstruct conditions in the oceans over the last 35,000 years. At times, we found that when benthic foraminifers indicate high organic content values at the bottom of the ocean this is not paralleled by high productivity signals at the sea surface, but instead must relate to changes in bottom-water circulation as a result of slower water circulation. The marine isotopic stages [MIS] 3-1 are clearly differentiated by benthic and foraminiferal assemblages as well as dinoflagellates and their cysts. MIS 3 is characterised by a much sharper [than today] thermocline that was closer to the sea surface and by the absence of a low-salinity barrier layer which today results from high monsoonal rains. The absence of the latter persisted during the last glacial period [MIS 2] when bottom circulation must have been reduced and high percentages of organic matter occurred on the sea floor combined with low dissolved-oxygen levels. The deglaciation is marked by a change in salinity at the sea surface as seen by the dinoflagellates and planktic foraminifers and progressive alteration of the thermocline was detected by foraminifers suggesting a less productive deep chlorophyll maximum in contrast with MIS 3 and 2. Monsoonal activity commenced around 15,000 cal years ago and was well established 2000 years later. The Holocene is marked by a significant increase in river discharge to the ocean, pulsed by the delivery of organic matter to the sea floor, despite overall oligotrophic conditions at and near the sea surface induced by a permanent low-salinity cap. We did not identify obvious and persistent upwelling conditions offshore Sumatra for the last 35,000 years. 2008 Elsevier B.V. All rights reserved. 1. Introduction 1.1. Regional oceanography and climate The eastern Indian Ocean adjacent to the 2 large Indonesian Islands, Sumatra and Java, displays contrasting seasonal characteristics at the sea surface. In the austral summer [December to February], salinity is rather low ( 32 S) due to the monsoonal activity. Overall, the monthly rainfall on Sumatra and offshore in the Indian Ocean is well over 30 cm per month [see http://www.jisao. washington.edu/data/smi/rss/precipssmirssclimr/jpg] with ensuing huge discharges of river water to the ocean. During that time, the predominant winds are such that the southeast trades, that are prevalent between December and April, will force the Southeast Java Current to flow in the direction of Western Australia, before turning southward to eventually join the Southeast Equatorial Current (Fig. 1). Corresponding author. Research School of Earth Sciences, The Australian National University, Canberra ACT 0200, Australia. Fax: +61 2 6125 5544. E-mail address: patrick.dedeckker@anu.edu.au (P. De Deckker). In contrast, during the austral winter, the northwest trade winds that are predominant between June and October cause the Current to run in the opposite direction to the one mentioned above, thus flowing towards the Andaman Sea. This coincides with the drier season, but still the Island of Sumatra continues to receive much rainfall. Consequently, salinity at the sea surface increases to 34.2. During the wet season there is a barrier layer (sensu Springtall and Tomczak, 1992) that will have been enhanced near the surface of the ocean and which consists of a low salinity cap that is approximately 50 to 100 m thick. The latter prevents any direct exchange between the atmosphere and the deeper ocean below. Of importance also is the Deep Chlorophyll Maximum [=DCM] which sits below the barrier layer, and which can only be exploited by those members of the plankton that can live in deeper waters. The barrier layer s characteristics are typical of the Indo Pacific Warm Pool [abbreviated herewith to Warm Pool ], a body of water that today encompasses the entire Indonesian Archipelago and that extends as far east as the Pacific side of Papua New Guinea, in the south just north of Australia and west to the eastern Andaman Sea. The Warm Pool consistently has surface waters 28 C (Yan et al., 1992) and a low salinity cap due to 0031-0182/$ see front matter 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.palaeo.2008.06.001

196 D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213 Fig. 1. Map showing the major oceanic current at the surface in the eastern Indian Ocean and the location of the cores mentioned in this paper. Abbreviations: ECC: Equatorial Counter Current; EGC = Eastern Gyral Current; IMC: Indian Monsoon Current; ITF = Indonesian Throughflow; LC = Leeuwin Current; SEC+South Equatorial Current; SJC = South Java Current; WA = West Australian Undercurrent. Currents adapted from Fig. 1 in Wijffels et al. (1996). the large amount of monsoonal rainfall and fluvial discharges to the sea. Contrasting wind directions, especially over different seasons, can cause some changes to the upper part of the water column, such as a migration of the thermocline, but elsewhere, such as offshore Java, seasonal upwelling can occur, e.g. the Java Upwelling System (see Wyrtki, 1962). Offshore the Island of Sumatra, the thermocline is relatively shallow (~90 m) throughout the year (Levitus, 1998), but no upwelling has been recorded offshore the Island of Sumatra due to the barrier layer. One critical issue about understanding the interactions between the oceanic atmospheric processes in the Warm Pool region, which plays an important role in global climate (see for review De Deckker et al., 2002), is to determine whether conditions in the atmosphere and surrounding seas have remained the same and, in particular, at a time when the world experienced much colder conditions and significant lower sea levels such as the Last Glacial Maximum, between 23,000 and 19,000 years ago (Yokoyama et al., 2000, 2001; De Deckker and Yokoyama, in press). For this purpose, we have chosen a gravity core taken offshore Sumatra, where today extremely high rainfall is recorded, and where seasonal upwelling is known to operate today. In order to interpret past environmental conditions that span the last 37,000 years of history, we have used a variety of microfossils extracted from the core in order to determine changes through time and identifying conditions, such as those at the sea surface [through foraminifers], at the bottom of the ocean [through benthic foraminifers], and also with respect to salinities at the sea surface and their relationship through river discharges [through the presence of specific dinoflagellates]. The latter will inform us on any major changes in precipitation that may have occurred through time, and this would reflect on past atmospheric conditions. This is rather important as there is still too little information available on hydrological changes that occurred in such an important region that is considered today to be one of the key area supplying fresh water to the global ocean. 1.2. Previous work in the region De Deckker and Gingele (2002) and Gingele et al. (2002) carried out extensive studies of core BAR9442 which was collected at a depth of 2542 m, some 50 km from the core studied here. These authors described conditions that occurred offshore Sumatra spanning the last 80 ka, but the core suffered from the lack of sufficiently wellpreserved calcareous microfossils because of the high abundance of the giant diatom Ethmodiscus rex which occurred in many parts of the core. The latter biogenic component prevented obtaining a good chronology based on AMS dates due to the paucity of foraminifera; only a stable isotope chronology could be obtained for that core. Gingele et al. (2002), using clay mineral assemblages, were able to determine changes in monsoonal activity registered by terrigenous input to the sea floor. In particular, these authors demonstrated that

D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213 197 during the glacial period, spanning 70 to 20 ka, strong northeasterly winds associated with the East Asian Winter Monsoon intensified the Indian Monsoon Current and the eastward flowing South Java Current. 1.3. Core bar 9403 The core we have chosen to study is crucially located on the western margin of the Indo-Pacific Warm Pool which, today, is considered to play an important role in the thermohaline circulation of the global ocean, as the waters offshore Sumatra often display low salinities due to heavy monsoonal rains and constantly high (N28 C) sea-surface temperatures. 2. Materials and methods 2.1. Deep-sea core BAR9403 Piston core BAR9403 was recovered offshore of Sumatra [5 49.20 S, 103 61.90 E] (Fig. 1), west of its southern extremity at a depth of 2034 m. The first 276 cm of this core were sampled every 5 cm. The top 105 cm of the core consist of light yellow clay. Below 105 cm, the sediment consists of greyish olive clay, with some levels (155 cm and 245 255 cm) where the sediment is sandier. This core is located in an area under the influence of the South Java Upwelling System and, therefore, faunal changes of planktic and benthic foraminifera and dinoflagellates, if related to any variation of the productivity at the sea surface, ought to provide information about modifications of past upwelling intensity/activity. For the extraction of planktic and benthic foraminifers, approximately 3 cm 3 of each sample was dried in an oven to obtain dry weight. After that, the samples were immersed in a dilute (3%) hydrogen peroxide solution until the clays had fully disaggregated, and the various fractions dried in an oven at 60 C after being washed with a gentle water jet through a 63 µm. Sampling for organic-walled dinoflagellate cysts (herein termed dinocysts' for the sake of brevity) was carried out at 4 cm intervals throughout the core to a depth of 240 241 cm. The standardized processing procedure for dinocysts is described in detail by de Vernal et al. (1993), and involves dissolution of a known volume of oven-dried sediment with 10% HCl followed by 40% HF. Where necessary, samples were disaggregated by placing them in an ultrasonic bath (b1 min to prevent damaging the cysts), and the remaining residue was then filtered through 120 and 10 µm sieves. Preparation techniques using acetolysis, strong oxidants, alkalis and hot acids were avoided to prevent bias from selective dissolution of the more susceptible cysts. 2.2. Stable isotopes and isotope chronology The δ 18 O record of planktic foraminifera was measured in order to produce an age model for the studied cores, while the δ 13 C isotopic signal from the benthic foraminifera was measured in order to acquire information about the deep-water circulation and/or the presence of organic matter at the sea floor. The species used for isotope analyses were: Cibidoides wuellerstorfi for benthic foraminifera, and Globigerinoides ruber (white variety) for planktic foraminifera. C. wuellerstorfi occupies an epifaunal microhabitat (Lutze and Thiel, 1988) and the isotopic composition of its test appears in equilibrium with the TCO 2 of the ambient deep water mass (Duplessy et al., 1984; Altenbach and Sarnthein, 1989). For this reason, the isotopic trend of this species reflects the one characterising the water at the sea floor (Duplessy et al., 1984). The methodology followed for the preparation of the samples for isotopic analyses is the same for the two types of foraminifera. A number of specimens sufficient to reach the minimum weight of material (180 μg) detectable by the mass spectrometer, were handpicked from each sample from the N250 µm fraction. Specimens were then washed in alcohol and placed in an ultrasonic cleaner for 5 s (up to 10 s when processing benthic foraminifera), in order to eliminate any contaminating residual adhering to the foraminifer test. Oxygen- and carbon-isotopic data obtained are reported in the usual δ notation, which is referred to the PeeDee belemnite (V-PDB) standard. Samples were calibrated against the National Bureau of Standards calcite (NBS-19), assuming values of δ 18 O V-PDB = 2.20 and δ 13 C V-PDB = 1.95. Analyses were conducted utilising a Finnigan-MAT 251 mass spectrometer at the Research School of Earth Sciences (RSES) at the Australian National University (ANU). The external errors were 0.05, for δ 8 O, and 0.08, for δ 3 C. 2.3. AMS dating of planktic foraminifera Up to 400 specimens of the planktic foraminifer G. ruber, following the common practice followed nowadays by micropalaeontologists to reach good confidence levels with counts, were picked for each level in the core [see Table 1] for AMS chronology conducted at the Australian Nuclear Science and Technology Organisation [=ANSTO]. The foraminifer samples were then placed in a scintillation vial in pure ethanol and allowed to be in an ultrasonic bath for 1 min to clear all disaggregating particles. The clean foraminifers were then washed in milliq water and then dried. Graphite targets were then made at ANSTO at the ANTARES AMS Facility (see Fink et al., 2004). The δ 13 C values quoted in Table 1 relate solely to the graphite derived from the fraction that was used for the radiocarbon measurement. The data supplied by ANSTO have been rounded according to Stuiver and Polach (1977). Radiocarbon dates were calibrated using the using the CALIB 5.1 program (Stuiver et al., 2006), and the Marine04 radiocarbon age calibration curve (Hughen et al., 2004). For the oceanic reservoir correction the estimate of the regional ΔR mean for NW Australia and Table 1 AMS radiocarbon analyses of samples from core BAR9403 showing also calibrated ages ANSTO codeozi Depth in core(cm) Species analysed δ13c per mill Percent Modern Carbo Conventional 14 C a Calibrated age BP pmc 1σ error yr BP 1σ error yr BP 1σ error 771 0 2 G. ruber 0.3 84.11 0.42 1390 45 889 785 970 772 25 26 G. ruber 1.2 46.78 0.22 6105 40 6489 6395 6587 773 55 56 G. ruber 1.2 32.92 0.22 8920 60 9553 9448 9653 774 65 66 G. ruber 0.9 30.41 0.17 9560 45 10,373 10276 10472 775 100 101 G. ruber 0.8 26.62 0.16 10,630 50 11,838 11675 12055 776 125 126 G. ruber 0.8 19.38 0.12 13,180 60 15,025 14882 15197 777 155 156 G. ruber 2.8 13.17 0.11 16,280 70 19,031 18924 19115 778 185 186 G. ruber 2.7 8.35 0.11 19,940 110 23,207 22914 23495 779 205 206 G. ruber -0.4 6.28 0.18 22,230 230 26,174 780 265 266 G. ruber 0.1 2.2 0.13 30,640 510 35,680 Calibrated age calculated using Bard (1998)'s polynomial.

198 D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213 Java of 67±24 (Bowman, 1985; Southon et al., 2002) was obtained from the CALIB Marine Reservoir Correction Database. All dates reported within this paper are in calibrated radiocarbon years. For the 2 samples dated in the lower portion of the core (at 205 206 cm and 265 266 cm), the polynomial developed by Bard (1998) was used to calibrate the ages. 2.4. Planktic foraminifers and statistical treatment of the data Counts of planktic foraminifera were made on splits of the N150 µm fraction to provide a base level for the ecological counts, removing small juvenile and possibly unidentifiable foraminifera. Each sample was split by an Otto-micro splitter until approximately 400 specimens were present in the final split. Planktic foraminifera were identified, at a species level, to reconstruct faunal assemblages through time. The species nomenclature used in this study follows the taxonomy of Saito et al. (1981). Principal component analysis was conducted on the planktic foraminiferal counts to find ecological groups within the data. The C2 program, version 1.4 (Juggins, 2003) was used to perform the statistical analysis. In accordance with other studies (Martinez et al., 1998), species abundance below 1% was excluded from analysis and raw counts were entered into the program to reduce error. 2.5. Dinoflagellates Total cyst concentrations (per gram of dry sediment) were calculated as follows: Total cysts=g ¼ E added XT cysts =Ecounted =Sdry weight : where E added is the number of exotic spore grains (Lycopodium sp.) added to the sample, T cysts is the total number of dinocysts counted in that sample, E counted represents the total number of exotic pollen grains counted in the sample, and S dry weight is the dry weight of sediment used (in grams). Calculated dinocyst concentrations using the marker-grain method provides an estimate within a standard deviation of 10% for a 0.95 confidence interval (de Vernal et al., 1987; Marret and de Vernal, 1997). The ratio of heterotrophic (H) to autotrophic (A) species is calculated as follows: H=A ratio ¼ H cysts= ðh cysts þ A cystsþ: If selective preservation has not greatly influenced the dinocyst assemblages, the H/A ratio can indicate productivity in the surface waters (e.g., Dale and Fjellså, 1994; van Waveren, 1993; Versteegh, 1994; Cho and Matsuoka, 2001; Roncaglia, 2004), as heterotrophic species (H-cysts) are found in high numbers in upwelling regions due to the abundance of suitable nutrients and prey (e.g., diatoms, which are usually the dominant organism in upwelling-related phytoplankton). However, some caution needs to be exercised in using this ratio for productivity purposes, as post-depositional transportation of cysts and selective preservation of species that are sensitive to aerobic decay (e.g. Zonneveld et al., 1997, 2001) can skew the true signature of the H/A ratio. It is imperative to consider these factors before drawing any conclusions regarding productivity. 2.6. Benthic foraminifera and statistical treatment of the data Specimens from the N150 µm fraction of each sample were isolated, counted and mounted on micropalaeontological slides. In order to compare the results of the study of benthic foraminifera from the core with the result of the analysis of the core-tops, the absolute number of individuals for each species was converted as the percentage of total foraminifera present in each sample. Species present with a percentage N2% in at least 1 sample were used for statistical analyses. Similar to the situation observed for the core tops, the specimens belonging to the genera Fissurina, Lagena, Lenticulina, Oolina and Parafissurina were present in many samples with high species diversity. For this reason, all the species belonging to these genera and used for statistical analysis were grouped together as Fissurina spp., Lagena spp., Lenticulina spp., Oolina spp. and Parafissurina spp. The program STATISTICA 6.0 [StatSoft Inc., STATISTICA (data analysis software system), version 6, www.statsoft.com] was used to perform Q mode Factor Analysis (Principal Components) on the species dataset of each core sample. 2.7. Benthic foraminifera accumulation rate (BFAR): applications and problems A linear relationship between the accumulation rate of benthic foraminifera (BFAR) and the amount of organic matter reaching the sea floor was outlined by Herguera and Berger (1991) and by Herguera and Berger (1994). BFAR has been used worldwide as a proxy to estimate variations of the past carbon-flux rate to the sea floor (Struck, 1995; Thomas et al., 1995; Loubere, 1996; Herguera, 2000; Diester- Haass and Zahn, 2001; Diester-Haass et al., 2002; Rasmussen et al., 2002;). For the eastern Indian Ocean, the selective dissolution of calcareous foraminiferal tests is excluded, as the lysocline is presently indicated at a depth of 2400 m, north of 15 S, and at a depth of 3600 m south of 15 S (Martinez et al., 1998), deeper than the sites studied here. Finally, since the BFAR is a function of the linear sedimentation rate (LSR), this parameter needs to be carefully determined. For the examined cores, LSR was calculated considering AMS measurements (see Section 2.3). The total number of foraminifera isolated in each sample was used to calculate the benthic foraminifera accumulation rate (BFAR), following the formula proposed by Herguera and Berger (1991): BFAR ¼ ðfþdðlsrþdðdbsþ n=cm 2 Kyr ; where F is the abundance of foraminifera (n/g), LSR is the linear sedimentation rate (cm/kyr) and DBS is the dry bulk sediment (g/cm 3 ) [g = grams of dry sediment]. 3. Results 3.1. Chronology for core BAR9403 based on AMS dates and oxygen isotopes All the AMS results are presented in Table 1. The oxygen isotope chronology for piston core BAR9403 is based on the δ 18 OofG. ruber in combination with that of C. wuellerstorfi (Fig. 2) with tie points linked to the SPECMAP chronology set by Martinson et al. (1993) and also verified by the AMS dates. The program performing timeseries analysis of Paillard et al. (1996) was used to calculate ages for all levels studied in the core. Note that we did not join all the points for the stable isotope analyses done on the benthic foraminifers as some levels had insufficient material for analysis. Nevertheless, the trends are clearly visible and are parallel to those of the planktic foraminifers. 3.2. Planktic foraminifera A total of 24,374 planktic foraminifer individuals were identified from core BAR9403. The relative abundance of planktic foraminifera shows clear intervals of abundance change paralleling each of the marine isotope stages recorded in core BAR9403 (Fig. 3). MIS 3 is characterised by Neogloboquadrina pachyderma (dextral) recording average abundances of 6%. The relative abundance of N. pachyderma (dextral) reduces to b2% near the Last Glacial Maximum [LGM] and does not recover during the Holocene. The

D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213 199 Fig. 2. Chronology of core BAR9403 determined by a combination of the δ 18 O composition of planktic and benthic foraminifers and AMS dates done on selected horizons in the core [for more information refer to Table 1 and text]. MIS1-3 refers to the Marine Isotopic Stages 1 to 3. The boundaries between the 3 stages were chosen following the procedures set in Martinson et al., 1993. occurrence of species Neogloboquadrina dutertrei remains fairly stable throughout the entire core but a 5% increase in abundance is seen during MIS 3 and at the Holocene/ MIS 2 transition. Pulleniatina obliquiloculata increases in abundance during MIS 3 and the Holocene with abundances N10%, compared to b10% during the LGM. Some minor species, within core BAR9403, only appear during MIS 3 and give further insight into differing conditions in the water column. For example, Turborotalia quinqueloba appears in the record with a peak of 2% in relative abundance during MIS 3. Other minor species of deep-water dwellers such as Globorotalia truncatulinoides and Globorotalia crassaformis also only appear in the record during MIS 3. The most obvious change in the relative abundance of planktic foraminifera in core BAR9403 is shown by Globigerina bulloides. High abundances of Ga. bulloides (26%) occur at approximately 12 ka and another peak of abundance (22.7%) occurs during the Holocene. This is compared to the periods from MIS 3 to the LGM where the relative abundance of Ga. bulloides is generally b10%. Species Gr. menardii also increases its relative abundance from b8% during MIS 3 to peak abundance during MIS 2 of 16% at approximately 14 ka. Comparatively, periods of increased abundance of Ga. bulloides and Globorotalia menardii coincide with periods of low abundance for species such as Gs. ruber and Gs. sacculifer. This is most noticeable after the LGM during the peak relative percentages of Ga. bulloides (see Fig. 3). Species Gs. ruber records its highest relative abundance of 22.8% during the Holocene but also records a high relative abundance of 20.5% around the 15 ka yr BP. Similarly, species Gs. sacculifer records its highest relative abundance of 17% during the Holocene but also records a relatively high percentage (15%) at ~16.5 ka. 3.3. Principal component analysis of planktonic foraminifera The Eigenvalues of the first five axes from principal component analysis are provided in Table 2. The variance within the species counts of core BAR9403 is explained by each axis, which decreases for each successive axis. In 56 observations, 28 variables were identified and analysed. The main variance is explained by Component 1 (Eigenvalue 1). However, relatively high variance is explained in the first four components (Fig. 4). Component 1, which explains 32.3% of variance, is dominated with positive scores of Ga. bulloides, Gr. menardii and N. dutertrei. Component 1 is also associated with negative scores from the species Pulleniatina obliquiloculata, Gs. sacculifer and Gs. ruber. Component 2, which accounts for 24.4% of the variance in core BAR9403, is dominated by N. pachyderma (dextral), N. dutertrei, P. obliquiloculata and Globigerinita glutinata. This component is also associated with strong negative responses by Ga. bulloides, Gs. ruber and Gs. sacculifer. Component 3 accounts for 14.7% of the variance and is dominated by the positive score of Gt. rubescens and the strongly negative responses by Gs. sacculifer, N. dutertrei and P. obliquiloculata. Component 4 accounts for 7% variance and is dominated by the high positive species score of Gs. sacculifer and also associated with high negative species scores of Gr. menardii, Gs. ruber, P. obliquiloculata and Gn. glutinata. Component plots (bi-plots) were also used to define ecological groups between the first two meaningful components. In the component plots (Fig. 5), the distance between the species relates to the similarity of the species and, therefore, the greater the angle between the vectors the greater the difference between the species. In addition, the vector or line length indicates the importance of the

200 D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213 Fig. 3. Composite diagram showing the percentage distribution against time in cal. years BP of the main planktic foraminifer species counted in core BAR9403. The shaded area implies the LGM.

D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213 201 Table 2 Eigenvalues for the first 5 PCA axes carried out on planktic foraminifers from core BAR9403 Axis Eigenvalue 1 0.323 2 0.244 3 0.147 4 0.07 5 0.063 species in explaining the variance of the species counts. The species scores of the first two components are displayed in the plots above and explain 57% of the variance within core BAR9403 and noted as (1-2) behind the group number. Group 1 (1-2) consists of species with positive scores in component 1 and 2. N. dutertrei is the principal species in Group 1 (1-2). The other primary species N. pachyderma (dextral) is separated by a significant angle from N. dutertrei, lying close to the Group 1(1-2) and Group 4 (1-2) boundary, thus reducing the correlation between these species (Fig. 4). Group 1(1-2) highest phase of abundance is during MIS 3 and, momentarily, at ~11 ka (see Fig. 3). Group 2 (1-2) consists of species with positive species scores in the first component but negative scores in the second component. The primary species are Ga. bulloides and Gr. menardii, with secondary species Gl. aequilateralis and Ga. falconensis. This group shows an increase of abundance of 25% from ~15 ka to ~6.5 ka as indicated by the relative abundance graph. Group 3 (1-2) contains species with negative scores in the first and second components. Principal species Gs. sacculifer and Gs. ruber are highly correlated due to the small angle separating their scores. The secondary species include Gl. calida and O. universa. The abundance of this group increases after 10 ka. Group 4 (1-2) contains species with negative scores in the first component and positive scores in the second component. The dominant species is P. obliquiloculata and associated with secondary species Gn. glutinata, Gr. truncatulinoides (dextral), Gt. rubescens, Gt. tenellus, Gs. conglobatus. The sample scores indicate this group dominates during MIS 3 between ~35 ka to 18 ka and increases its abundance again after 9 ka. 3.4. Dinoflagellates Dinocyst concentrations ranged from between 616 and 12,231 cysts/ g for the last ~31 ka in the BAR9403 core, although only 3 horizons recorded concentrations in excess of ~3300 cysts/g (9.5 ka with 8230 cysts/g; 5.3 ka with 11,215 cysts/g, and 0.9 ka with 12,231 cysts/ g) (see Figs. 6 7). Twenty-three species of dinoflagellate cysts belonging to the gonyaulacoid (all autotrophic) and protoperidinioid (all heterotrophic) groups were identified in this core, of which Brigantedinium and Spiniferites dominated. From 30.8 to 23.9 ka (late MIS 3), concentrations fluctuated between 1489 and 820 cysts/g. Through this interval, the assemblages were dominated by heterotrophic species (predominantly from the genus Brigantedinium), with the H/A ratio remaining consistently above 0.57 and ranging up to 0.84. There is a noticeable decrease in cyst concentrations before the onset of the LGM, from 1489 cysts/g at 26.6 ka, then at the start of the LGM with 1010 cysts/g at 23.9 ka to 674 cysts/g at 22.5 ka. A sudden increase in concentrations occur at 21.8 ka with values more than doubling to 1585 cysts/g. However, concentrations then show a steady and relatively consistent decrease, from 957 cysts/g at ~21.1 ka to ~616 cysts/g at 15 ka, although there are two small peaks at 17.7 ka (930 cysts/g) and 15.7 ka (913 cysts/g). Despite these minor peaks, the concentrations through this period are some of the lowest recovered from the whole core. At the same time, the H/A ratio generally remained between 0.76 and 0.64 for most of this interval. This decreasing trend in cyst concentrations is broken by a threefold increase at ~14.4 ka (1840 cysts/g), with concentrations subsequently staying above ~1400 cysts/g throughout the remainder of the core (see Fig. 8). Relative peaks in cyst concentrations occurs every ~1.6 ka from ~17.7 ka onwards. O. centrocarpum virtually disappears from the core after ~26.7 ka, being present only at concentrations of b22 cysts/g at 21.1 ka, 19 ka and 17 ka. This species becomes more (relatively) dominant from ~14.4 ka until ~9.9 ka, although it is absent at 11.8 ka and at 11 ka. Concentrations do not increase past 230 cysts/g at any stage, however. Minor reversals from a heterotrophic-dominated assemblage to an autotrophic one are identified at 19.7 ka (354 cf 495 cysts/g: H/A ratio 0.42), 17 ka (364 cf 380 cysts/g: H/A ratio 0.42) and 13.1 ka (690 cf 842 cysts/g: H/A ratio 0.45). However, autotrophic species (predominantly species of Spiniferites) grow frequently more dominant than heterotrophic species from 10.8 ka onwards. The previously unrecorded Spiniferites 1 is only present in the upper two samples (5.3 ka and 0.89 ka [core top]) from BAR9403, although concentrations remain low (156 and 117 cysts/g respectively). 3.5. Benthic foraminifers 3.5.1. Factor analysis Ninety-three benthic foraminifera species were identified among the BAR9403 samples. Mean species percentages ranged between 17.8% (Bulimina aculeata) and 0.04% (Siphogenerina raphanus). The taxa showing a percentage N2% in at least one sample were selected for Q- mode Factor analysis (Principal Components). Factor analysis calculated four varimax factors, which explained 85.35% of the variance of the species distribution. The species scores are listed in Table 3. The four factors are dominated by four taxa all characterised by high mean percentages. F1 is dominated by B. aculeata (43), F2 is dominated by O. t. umbonatus (51), F3 is dominated by Epistominella exigua (44) and F4 is dominated by Cibicidoides wuellerstorfi (53) (number of occurrence is given in brackets) (Fig. 9). Dominant species percentages are shown in Fig. 10. Oridorsalis tener umbonatus showed high percentages from 35 to 26 kyr BP (termination of MIS3 MIS2), while values between 10% and 20% were recorded between 12 and 6 ka (termination of MIS2 late MIS1). B. aculeata showed high percentages (N30%) from 26 until 14 ka (MIS2). Between 9 and 4 ka (late MIS1), this species was characterised by percentages ranging between 10% and 20%. E. exigua reached percentages N10%, at 35 ka, between 24 and 18.5 ka (early MIS2) and during the last 6 kyr. C. wuellerstorfi was characterised by high relative percentage between 14 and 6 ka (termination of MIS2 late MIS1). This taxon reached percentages N10% between 35 and 26 ka (mid MIS3). 3.5.2. Faunal characteristics Between 35 and 15 ka (MIS3 termination of MIS2), the agglutinated species percentages were always b5%. They were characterised by high percentages for the last 15 kyr (termination of MIS2 MIS1), between 3% and 14.5%. Porcellaneous taxa percentages before 27 kyr BP were N15%. Between 27 and 15 kyr BP (mid MIS3 mid MIS2) values decreased, being b15%, and increased again during the last 16 kyr. Infaunal taxa were b40% before 27 ka. Between 27 and 13 ka, percentages were always N50%, with a peak of 68% (at 17 ka). During the last 10 kyr, this group of species displayed percentages b60%, ranging between 28% and 57%. Diversity indices α, H(S) and E followed similar patterns, steadily increasing from 35 kyr BP to the Present. A major limit can be found at 15 ka: after this point, until the Present, α and E displayed values ranging between 15 and 25; H(S) displayed values generally above 2.7. Dominance (D) showed high values between 25 and 14 ka.

Fig. 4. Plot of the relative abundance of the four groups as found by the principal component analysis of planktic foraminifer from core BAR9403 plotted against the chronology in cal. years BP as well as the stable isotope record for oxygen and carbon obtained on the planktic species G. ruber. Note the 3 MIS indicated in the right hand column. The shaded area implies the LGM. 202 D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213

D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213 203 Fig. 5. Bi-plot of Components 1 and 2 for planktic foraminifer counts that explain 58% of the variance in core BAR9403 and which also reveal four ecological groups [see text for more information]. 3.5.3. Benthic Foraminifera Accumulation Rate (BFAR) and accumulation rates calculated for B. aculeata, E. exigua and Uvigerina proboscidea The selective destruction of soft-cemented agglutinated foraminifera is instead a phenomenon recorded for the analysed samples, but it only affects a minimal percentage of the total assemblage and should not preclude the use of BFAR. The BFAR curve (see Fig. 10) showed values ranging between 71 and 240 n/cm 2 kyr 1, for the period between 35 and 29 ka. Between 29 and 14 ka, BFAR was generally higher, reaching values N250 n/cm 2 kyr. After 15 ka, BFAR was characterised by lower values. Between 14 and 10 ka, BFAR decreased passing from 180 to 90 n/cm 2 kyr and between 10 and 6 kyr BP, it ranged between 160 and 53 n/cm 2 kyr. During the last 6 kyr, BFAR decreased, reaching a value of 20 n/cm 2 kyr 1 for the Present. The accumulation rate of B. aculeata reached values N30 n/cm 2 kyr 1, between 29 and 14 ka, with a peak of 170 n/cm 2 kyr 1 at 22 ka. After 14 ka, this species was nearly absent. A small increase in B. aculeata AR was recorded for the last 6 kyr. E. exigua was characterised by high AR, between 29 and 14 ka, ranging between 10 and 60 n/cm 2 kyr 1. After 14 ka, the AR of this taxa was nearly 0. It increased again during the last 6 kyr, reaching values N10 n/cm 2 kyr 1. U. proboscidea followed a pattern similar to the former two species, although it was characterised by lower AR values compared to the other two taxa. Between 29 and 14 ka, U. proboscidea AR ranged between 5 and 20 n/cm 2 kyr 1. After 14 ka, this species displayed AR values b5 n/cm 2 kyr 1, with an isolated peak at 9ka,whenitreachedanARof15n/cm 2 kyr 1. 4. Discussion 4.1. Evidence from planktic foraminifers The distribution of planktic foraminifera in core BAR9403 implies phases of variation within the mixed layer since MIS 3. Overall, MIS 3 appears to be a period of reduced vertical mixing resulting in the stratification of the mixed layer. PCA analysis (see Fig. 5) and the sample scores indicate that Group 1 (1-2), containing species Neogloboquadrina dutertrei and Neogloboquadrina pachyderma, dominated over the other ecological groups during MIS 3 and momentarily at 11 ka (75 cm). Species N. dutertrei and N. pachyderma are both considered thermocline dwellers whose increased abundance has been linked to the development of a Deep Chlorophyll Maximum layer (=DCM). A DCM is initiated when the upper water column is stratified and an increase in chlorophyll production occurs at the thermocline (Fairbanks and Weibe, 1980; Thunell and Sautter, 1992), therefore, resulting in optimal conditions for these heterotrophic species. Cluster analysis by Ding et al. (2006) of core-tops within the Indonesian Archipelago classified the surface waters of the Sumatra Region as being relatively oligotrophic, low in salinity, small in seasonal temperature differences, shallow thermocline and of high dissolution. Ding et al. (2006) revealed that N. dutertrei obtained the second highest abundances in the Indian Monsoon Sumatra Region where core BAR9403 is located, due to a shallow thermocline and low salinity waters. In the relative abundance counts in core BAR9403, N. dutertrei is the most abundant species throughout the entirety of the core with an increase in abundance of 5% during MIS 3 and 11 ka. However, N. pachyderma (d) is not present in Ding et al (2006) coretop counts. The appearance of N. pachyderma (d) species during MIS 3 in core BAR9403 indicates a reduction in temperatures at the base of the thermocline compared to the present. High frequencies of N. pachyderma are linked to waters with low temperature (10 14 C), low salinity (34 and 35.5), and high nutrient (phosphate) levels 1.7 µg at/l (Bé and Hutson, 1977). It is also shown through the abundances of Globigerina bulloides and Globorotalia menardii, the principal species of PCA Group 2 (1-2) that upwelling conditions did not occur during MIS 3 (Fig. 3) but, instead, the conditions favoured the development of a DCM layer. Species, Ga. bulloides and Gr. menardii, are both considered to be upwelling indicators in tropical regions (Bé, 1977; Thunell and Reynolds, 1984; Auras-Schudnagies et al., 1989; Kroon and Nederbragt, 1990; Martinez et al., 1998; Ganssen and Kroon, 2000). Our conclusion relies on the fact that Group 2 (1-2) does not show a

Fig. 6. Plot of the major dinoflagellate species and cysts recovered in core BAR9403 against time in cal. years BP. Note that the distribution of taxa does not extend as far back in time as for the foraminifers in the same core as no dinoflagellates were recovered below 31 ka. The shaded area implies the LGM. 204 D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213

D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213 Fig. 7. Plot of the minor dinoflagellate species recovered in core BAR9403 against time in cal. years BP. Note that the distribution of taxa does not extend as far back in time as for the foraminifers in the same core as no dinoflagellates were recovered below 31 ka. The shaded area implies the LGM. 205

206 D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213 Fig. 8. Plot of autotrophic (A) and heterotrophic (H) dinoflagellate species and the H/A ratio that is used to indicate productivity at the sea surface. The shaded area identifies the LGM. response to the existence of colder water and possible nutrients at the sea surface during MIS 3 (Fig. 3). This again suggests that the mixed layer was stratified and nutrients did not reach the sea-surface and, therefore, upwelling conditions were not established during MIS 3. De Deckker and Gingele s (2002) analyses of core BAR9442, located ~50 km from BAR9403, indicated blooms of the giant diatom Ethmodiscus rex between 28 ka and 19 ka. These authors claim that blooms occur when the water-column is permanently stratified, with a substantial increase in salinity and high levels of silica and nitrate near the sea-surface. This period coincides with the indicated phase of stratification within core BAR9403. In addition, a regional pattern of DCM development was already observed within the various seas of the Throughflow region by Linsley et al. (1985), Barmawidjaja et al. (1993), Ding et al. (2002) and Spooner et al. (2005) for MIS 3 and MIS 2. The primary species of Group 4 (1-2) is Pulleniatina obliquiloculata. The sample scores indicate this group dominates during MIS 3 between ~28 ka to 23 ka and increases its abundance again at ~19 ka. Pulleniatina obliquiloculata is predominantly a tropical species and has been associated with warm-water masses such as the South Equatorial Current (Bé and Hutson, 1977) and the Kuroshio Current (Wang et al., 1999; Li et al., 1997).

D.S. Murgese et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 265 (2008) 195 213 207 Table 3 Factor scores of the Q-mode varimax factors in piston core BAR9403 Species Factor 1 Factor 2 Factor 3 Factor 4 Allomorphina pacifica 0.3089 0.1891 0.1440 0.4228 Anomalina globulosa 0.1029 0.3378 0.5053 0.1177 Astrononion echolsi 0.0757 0.2354 0.5619 0.2702 Bolivina robusta 0.1284 0.2581 0.4080 0.3951 Bolivinita quadrilatera 0.0449 0.2518 0.2155 0.4513 Brizalina semilineata 0.2288 0.2768 0.3201 0.3363 Bulimina aculeata 7.2886 0.5376 1.0540 0.6130 Bulimina costata 0.0036 0.8853 0.8493 0.5370 Bulimina exilis 0.1731 0.3728 0.3462 0.7523 Cassidulina laevigata 0.1427 0.7604 0.1672 0.1645 Ceratobulimina pacifica 0.1552 0.2852 0.3917 0.4301 Chilostomella oolina 0.1833 0.0229 0.5445 1.1802 Cibicidoides bradyi 0.1231 0.2956 0.1109 0.8635 Cibicidoides pseudoungerianus 0.1755 0.4268 0.3179 0.2755 Cibicidoides robertsonianus 0.2503 0.3103 0.0965 0.4739 Cibicidoides wuellerstorfi 0.4494 0.0754 0.0115 5.6760 Eggerella bradyi 0.1987 0.3063 0.0625 0.3695 Epistominella exigua 0.9489 1.0216 6.7501 0.9394 Fissurina sp. 0.0773 0.2433 0.0691 0.1458 Fursenkoina bradyi 0.1916 0.2707 0.2176 0.5379 Fursenkoina fusiformis 0.0744 0.3702 0.2660 0.2612 Fursenkoina sp. 0.2311 0.3101 0.2070 0.4578 Gavelinopsis lobatulus 0.0627 0.3006 0.4778 0.1508 Globobulimina affinis 0.2283 0.1015 0.4461 0.3164 Globobulimina pacifica 0.2208 0.3368 0.3832 0.0175 Globocassidulina subglobosa 0.1451 0.4291 0.0162 0.7227 Gyroidinoides orbicularis 0.2249 0.3363 0.4198 0.6359 Gyroidinoides polius 0.2782 0.3311 0.1473 0.1638 Gyroidinoides soldanii 0.2196 0.1538 0.4720 0.3148 Hoeglundina elegans 0.2484 0.2048 0.3849 0.7204 Hyalinea balthica 0.2711 0.2094 0.0404 0.4843 Karreriella bradyi 0.2885 0.2768 0.0969 0.3526 Lagena sp. 0.2595 0.3551 0.0259 0.2310 Lenticulina sp. 0.2639 0.1505 0.2030 0.4846 Loxostomum karrerianum 0.2686 0.2785 0.0740 0.4608 Melonis barleeanum 0.4038 0.0621 0.0222 1.6341 Melonis pompilioides 0.2687 0.2924 0.1615 0.4811 Miliolinella subrotunda 0.2506 0.1645 0.2867 0.5811 Nonionella bradyi 0.2498 0.1729 0.3645 0.4611 Nummoloculina irregularis 0.2253 0.1223 0.3162 0.5907 Oridorsalis tener umbonatus 0.6304 7.0328 0.6723 0.0899 Osangularia cultur 0.1744 0.1194 0.4819 0.3035 Parafissurina sp. 0.1627 0.2137 0.3682 0.4435 Pullenia bulloides 0.2492 0.3133 0.2096 0.0141 Pullenia quinqueloba 0.2011 0.3257 0.0019 0.1965 Pyrgo depressa 0.1689 0.0409 0.4007 0.4149 Pyrgo lucernula 0.2457 0.0284 0.4646 0.6110 Pyrgo murrhina 0.0549 0.9087 1.6953 0.4588 Pyrgo serrata 0.1001 0.1312 0.3007 0.5876 Quinqueloculina seminulum 0.2555 0.1375 0.2256 0.3546 Quinqueloculina venusta 0.1772 0.7055 0.3634 0.8163 Robertinoides brady 0.1222 0.2627 0.3583 0.5413 Sigmoilopsis schlumbergeri 0.1801 0.0502 0.7339 0.8957 Siphogenerina raphanus 0.2631 0.1826 0.2915 0.5857 Siphotextularia catenata 0.2008 0.0380 0.3727 0.5857 Uvigerina peregrina 0.0874 0.5457 0.8255 1.5660 Uvigerina proboscidea 0.3117 0.3550 0.5466 2.3568 Valvulinaria araucana 0.2753 0.1747 0.0729 0.4287 The species listed in bold are the 4 taxa whose distributions are presented in Fig. 10. The sample scores reveal that Group 4 (1-2) and Group 1 (1-2) alternate in dominance during MIS 3. Previous studies have linked the abundance of P. obliquiloculata to its sensitivity to winter temperatures and, therefore, high concentrations of P. obliquiloculata occur during warm winter temperatures at the subsurface (Li et al., 1997; Pflaumann and Jian, 1999). The high relative abundance of P. obliquiloculata within Group 4 (1-2) and N. pachyderma in Group 1 (1-2) during MIS 3 initially appears to be contradictory but may also support the hypothesis that the water column was stratified during this period. The study by Bé (1977) in the region of core BAR9403 found P. obliquiloculata predominantly between 50 and 100 m while N. pachyderma was predominantly below 100 m. Either stratification provided temperature niches for both these species or the temperature at the base of the thermocline went through seasonality phases. Alternatively the deepening of the mixed layer into cooler, deeper waters could explain the increased abundance of N. pachyderma (d) during MIS 3. Thickening of the mixed layer due to stratification has been observed within the Santa Barbara Basin by Pak and Kennett (2002). Pulleniatina obliquiloculata is one of the most dissolution-resistant species in the low latitudes and can make up to 70% of the total fauna in the Western Pacific sediments(thompson, 1981 and Thiede et al., 1997). It is believed that dissolution is not the controlling factor on the abundances of planktic foraminifera in core BAR 9403 due to the lack of etching on the foraminifer tests. In addition, the low relative percentages of other dissolution-resistant species such as Gr. menardii during MIS 3 and the Holocene coincide with high abundances of P. obliquiloculata. During MIS 2, the relative abundance of N. dutertrei reduced slightly, N. pachyderma relative abundance reduced to b1% and the low abundance of upwelling species such as Ga. bulloides was maintained from MIS 3 to MIS 2 (see Fig. 3). The PCA analysis grouped Gs. ruber and Gs. sacculifer together and the sample scores indicated that Group 3 (1-2) dominated over the other PCA groups during the LGM and the Holocene. From these results the LGM appears to be characterised by the abundance of tropicalsubtropical species with a preference for oligotrophic conditions. In addition, the higher abundance of these tropical-subtropical species suggests the water column was still stratified and nutrients may not have been entrained into the upper surface layer through wind-forced mixing which occurs presently (Sprintall et al., 2002). The major change registered in core BAR9403 is around the upwelling signal of PCA Group 2 (1-2) containing Ga. bulloides and Gr. menardii. The sample scores indicate that this group dominated from 14 ka to 9 ka (Fig. 3). Consequently, we interpret this phenomenon to indicate that nutrients were upwelled to the sea-surface for that time period. The abundance of Group 3 (1-2) and Group 4 (1-2) was also reduced from 15 ka to 9 ka and this implies the removal of the stratified structure of the water column. At the LGM-Holocene transition, De Deckker and Gingele (2002) recorded a high Ba excess, and low Erex abundance in core BAR9442 and suggested the removal of stratified conditions and this further supports our interpretation for the region. Presently, upwelling along the coast of Java operates during the SE Monsoon as a result of southward Ekman transport (Wyrtki, 1962) but nutrients do not reach the surface due to an increase in the transport of the Throughflow (Bray et al., 1997). Martinez et al. (1998) showed that the change in abundance of species, such as N. pachyderma, N. dutertrei and Ga. bulloides, varies around the region and depend on mixed layer thickness and the intensity of the Java Upwelling System. A more dynamic and productive Java Upwelling System encouraged by the SE Monsoon during the LGM is indicated in studies by Martinez et al. (1999), Takahashi and Okada (2000), Gingele et al. (2001) and Gingele et al. (2002). In addition, palaeoproductivity proxy investigations by Müller and Opdyke (2000) in the Timor passage indicate increased nutrients in the mixed layer during MIS 2. However, results from cores BAR9403 and BAR9442 off the coast of Sumatra suggest that upwelling did not occur until after the LGM. It is possible the upwelling signal in core BAR9403 is evidence of a more intense Java Upwelling System and hence monsoonal system from 14 ka. 4.2. Evidence from dinoflagellates 4.2.1. Possible influences upon cyst assemblages Before any worthwhile interpretation of the dinocyst fossil record from BAR9403 can be made, it is imperative that post-depositional processes such as transportation, selective decay and dissolution are considered, as these factors can create a biased view of the actual thanatocoenosis. We believe that the effect of sea level changes through time did not affect the abundance of dinoflagellate cysts per se as there was little change of the coast line with respect to the core